From ICE Age to rEVolution

The Determinants of Green and Greenish Innovations in the Automotive Industry


Marco Guerzoni

(with N. Corrocher, A. Novaresio, T. Pierangeli)

UNIVPM Seminar – 13/02/2025

Introduction

  • Transport accounts for ~10% of global CO2 emissions.
  • Innovation in green automotive technologies is crucial for sustainability.
  • EU has implemented policies to support EV adoption.
  • However EU and OECD countries have stepped back from competing with China in the green automotive sector.
  • Key drivers of innovation: Demand, Technology, and Regulation.

Ursula von der Leyen: “Europe must lead the green transition, and mobility is at the core of this transformation.”

Early Competition in the Automobile Industry

  • Rapid Growth: The automobile industry expanded quickly in the 1890s.
  • Competing Technologies: Electric, steam, and gasoline cars all vied for market share.
  • Market Trends: Although U.S. developments lagged Europe, steam cars (e.g., the Locomobile) were initially the most popular.
    Locomobile Steam Roundabout 1900

    Locomobile Steam Roundabout 1900

The Promise of Electric Vehicles

  • Strong 1899 Sales: In 1899, 1,575 electric vehicles were sold alongside 1,681 steam and 936 gasoline cars.
  • Strategic Moves: The Electric Vehicle Company ordered 200 vehicles and announced plans for large-scale electric taxicabs.
  • Technical Advantages: Electric vehicles shared components with streetcars and showcased advanced performance (e.g., “La Jamais Contente” reached 100 km/h).

The Promise of Electric Vehicles

  • Strong 1899 Sales: In 1899, 1,575 electric vehicles were sold alongside 1,681 steam and 936 gasoline cars.
  • Strategic Moves: The Electric Vehicle Company ordered 200 vehicles and announced plans for large-scale electric taxicabs.
  • Technical Advantages: Electric vehicles shared components with streetcars and showcased advanced performance (e.g., “La Jamais Contente” reached 100 km/h).
Jamais Contente 1899

Jamais Contente 1899

ICE: dominant paradigm and lock-in

  • Rising Competition: Despite a doubling of electric vehicle sales from 1899 to 1909, gasoline car sales increased more than 120 times.
  • Market Shift: The rapid expansion of gasoline technology, bolstered by factors like patent control and economies of scale, ultimately sidelined electric vehicles.
  • However: Electric vehicle technologies never really disappeared, but survived in the market itself, in adjacent markets, and in different markets (e.g. battery for renewable and mobiles)
  • Key Research Question: How do we account for the present renewed competition between different standards, both theoretically and empirically?

Eco-Innovation and the Automotive Sector

  • Definition: Eco-innovations include new or modified products, processes, practices, systems, and institutional arrangements that benefit the environment and sustainability (Díaz-García et al., 2015; Rennings, 2000; Oltra et al., 2008a).
  • Sector Significance: The automotive industry is a major contributor to greenhouse gas emissions and urban air pollution (Nunes & Bennett, 2008).
  • Global Targets: Achieving net zero by 2050 (IEA, 2021; Bouckaert et al., 2021) requires phasing out new internal combustion engine (ICE) passenger cars by 2035.

Historical & Technological Context

  • Environmental Awareness: The 1970s oil shocks and increasing environmental awareness spurred efforts to “green” the engine (Meadows et al., 1972; Faria & Andersen, 2017).
  • Innovation Debates: Although electric vehicles (EVs) lower local emissions, they also pose challenges (e.g., battery-related issues, cost, autonomy) (ADL Report, 2016; Cecere et a. 2018).
  • Industry Focus: Research highlights competing technological trajectories in the automotive sector, emphasizing product design and eco-innovation (Segarra-Ona et al., 2011; Sierzchula et al., 2012; Christensen, 2011; Mazzei et al., 2023).

Lock-In, Path Dependence & Technological Inertia

  • Lock-In Concept: “Carbon lock-in” occurs when investments in ICE technologies create inertia against greener alternatives (Unruh, 2000; Unruh, 2002).
  • Three Lock-In Sources: Demand, supply, and regulatory lock-ins reinforce the dominance of ICE technology (Dijk & Yarime, 2010; Bjørnavold & Van Passel, 2017).
  • Industry Example: The automotive cooling systems in the EU show how even well-intended regulations can inadvertently entrench existing technologies (Bjørnavold & Van Passel, 2017).

Escaping the Lock-In: Regulation, Technology & Demand

  • Regulatory Drivers: Stringent and targeted environmental policies (e.g., fuel efficiency programs, emission standards) encourage eco-innovation (Kuik, 2007; Maldonado-Guzmán et al., 2024; Haščić et al., 2008).
  • Technology Push: Collaborative R&D and supplier partnerships foster the development of new automotive technologies, including electric, hybrid, and fuel cell innovations (Potter & Graham, 2019; Wu et al., 2020; Scarpellini et al., 2012).
  • Demand Pull: Market demand and consumer preferences for sustainable products drive eco-innovative efforts (Fernando et al., 2021; Peiró-Signes & Segarra-Ona, 2018).

Emerging Trajectories: Greenish vs. Green Innovations

  • Two Trajectories:
    • Greenish Innovations: Incremental improvements aimed at enhancing ICE efficiency.
    • Green Innovations: Radical eco-innovations such as hybrid, electric, and fuel cell technologies (Aghion et al., 2016; Mazzei et al., 2023).
  • Technological Competition: The “sailing ship effect” and “red queen effect” illustrate how incumbent technologies may persist even amid radical innovations (Sick et al., 2016; Barnett & Hansen, 1996; Derfus et al., 2008).
  • Key Drivers: Shifts in global market demand, changes in fueling infrastructure, and climate policies (Dijk et al., 2013) are critical in determining which trajectory prevails.

Conclusion in the literature

  • Mixed Outcomes: Eco-innovation drivers may either disrupt the dominant ICE paradigm or reinforce it by “greenishing” existing technologies (Dijk et al., 2013; Bergek et al., 2014)
  • Policy Implications: An integrated approach combining regulation, R&D investments, and market incentives is essential to overcome path dependency (Costantini et al., 2015; Demirel & Kesidou, 2019; Cabigiosu & Lanzini, 2023).
  • Future Gaps: Further empirical research is needed to assess the interplay of public policy, technological breakthroughs, and market demand in driving the diffusion of low-emission vehicles (Mazzei et al., 2023; Oltra & Saint Jean, 2009).
  • In this paper: We test the correlation of technological factor, demand effect and regulation upon the two current greenish and green trajectories.

Data and Methodology

  • Panel data from OECD countries (2005-2021), supplemented with OICA (2022) and IEA (2022)
  • Key variables:
    • Dependent: Green & Greenish patent counts.
    • Independent: EPS, R&D investments, EV demand, GDPpc, Technological spezialization
    • Controls: Fuel prices, material footprint.

Handling Missing Data

  • Missing values accounted for 5.7% of the dataset.
  • Multiple Imputation by Chained Equations (MICE) used.
  • Verified Missing At Random (MAR) assumption.

Green and Greenish Patents

Variables in dataset

Variables in dataset

Evolution of Green and Greenish patents

Evolution of EPS over time 1/2

Evolution of EPS over time 2/2

Model Selection

  • Two-way fixed effects model with Poisson estimation.
  • It has been preferred over alternatives such as OLS-panel with log patent.
  • 2 years lagged variables.
  • Pseudo-demeaning.
  • Accounting for non-linear effects of EPS, R&D, and GDP.
  • Accounting for time and country fixed effects.

Results

Variabile Green Greenish
EPS -0.502*** 0.045**
EPS^2 0.075*** -0.068**
Log Green/ishPat 0.097*** 0.030***
Log Green/ishPat^2 0.40** ns
Relative Tech Advantage 2.496*** 0.040***
Relative Tech Advantage^2 -0.823** 2.273***
Green Vehicle % 1.987*** -0.656**
Material Footprint -0.019 -1.355**
GDP per Capita 0.079*** ns
GDP per Capita^2 -0.001*** ns

Results: EPS

Results: Patents, RTA, and GDP

Results: Fixed effects

Robustness Checks

  • Tested model with different lag structures.
  • Splitting sample into high vs. low patent-producing countries.
  • Bayesian Generalized Additive Models (GAMs) confirm results.

Results

  • EPS pushes GREEN only at high level, otherwise the opposite.
  • There are spillovers between different paradigms
  • EV demand pulls GREEN technologies and crowd-out GREENISH efforts
  • Excess of specialization strenghen the ICE lock-in, but some is good.

Thanks

$ echo 
marco.guerzoni@unimib.it

$ echo 
Feel free to contact me for research collaborations or discussions.

$ ping -c 1 127.0.0.1
PING 127.0.0.1 (127.0.0.1): 56 data bytes
64 bytes from 127.0.0.1: icmp_seq=0 ttl=64 time=0.045 ms

$ echo "Bicocca is home."

The Fixed Effects Challenge

  • Standard linear panel model with fixed effects: \[ Y_{it} = X_{it}\beta + \alpha_i + \varepsilon_{it} \]
  • Two estimation approaches:
    1. LSDV (dummy variables)
    2. Within transformation (demeaning)

From Full to Pseudo-Demeaning

  • Traditional within transformation: \[ Y_{it} - \bar{Y}_i = (X_{it} - \bar{X}_i)\beta + (\varepsilon_{it} - \bar{\varepsilon}_i) \]
  • Pseudo-demeaning transformation: \[ Y_{it}^* = Y_{it} - \theta_i\bar{Y}_i, \quad X_{it}^* = X_{it} - \theta_i\bar{X}_i \] where θi is the scaling factor

The Basic Setup

  • Standard Poisson with two-way FE: \[ E[y_{it}|x_{it}] = \exp(x_{it}\beta + \alpha_i + \gamma_t) \] where:
    • αi: unit fixed effects
    • γt: time fixed effects

Conditional Likelihood

  • Joint probability: \[ P(y_{i1},...,y_{iT}|\sum_{t=1}^T y_{it}) = \frac{\prod_{t=1}^T \exp(x_{it}\beta + \alpha_i + \gamma_t)^{y_{it}}}{\sum_{d_i \in D_i} \prod_{t=1}^T \exp(x_{it}\beta + \alpha_i + \gamma_t)^{d_{it}}} \]

The CML Trick

After conditioning on Σt yit: \[ P(y_{i1},...,y_{iT}|\sum_{t=1}^T y_{it}) = \frac{\prod_{t=1}^T \exp(x_{it}\beta)^{y_{it}}}{\sum_{d_i \in D_i} \prod_{t=1}^T \exp(x_{it}\beta)^{d_{it}}} \]

  • Fixed effects αi and γt disappear
  • β can be estimated consistently
  • No incidental parameters problem

Ex-Post Fixed Effects Recovery

Retrieving Unit Fixed Effects

For each unit i: \[ \hat{\alpha}_i = \ln\left(\frac{\sum_{t=1}^T y_{it}}{\sum_{t=1}^T \exp(x_{it}\hat{\beta} + \hat{\gamma}_t)}\right) \]

Retrieving Time Fixed Effects

For each period t: \[ \hat{\gamma}_t = \ln\left(\frac{\sum_{i=1}^N y_{it}}{\sum_{i=1}^N \exp(x_{it}\hat{\beta} + \hat{\alpha}_i)}\right) \]

Implementation in R

Estimation with fixest

# Two-way FE Poisson
mod_twfe <- feglm(y ~ x | id + time, 
                  family = "poisson",
                  data = panel_data)

# Retrieve FE
alphas <- fixef(mod_twfe)$id    # Unit FE
gammas <- fixef(mod_twfe)$time  # Time FE

Standard Errors

  • For recovered fixed effects: \[ SE(\hat{\alpha}_i) = \sqrt{\frac{1}{\sum_{t=1}^T \exp(x_{it}\hat{\beta} + \hat{\gamma}_t)}} \]
# Compute SE
se_fe <- se(mod_twfe, 
            vcov = "cluster",
            cluster = ~ id)